AI Agent

Demo

Live interaction flow of the chat UI with the current deployed environment.

Overview

AI Agent is a general AI portfolio project with a chat interface and an integration-first backend. The app sends user prompts to an n8n webhook flow and returns the response in a simple chat UX. Interactions and execution metadata can be persisted in Supabase for observability.

Architecture

UI (public/) -> POST /api/agent -> rate-limit middleware -> controller -> service -> n8n webhook
                          \-> saveInteractionLog -> Supabase
UI (public/) -> GET /api/agent/metrics -> controller -> Supabase aggregate (fallback: runtime)

API

POST /api/agent

Send a user message to the assistant.

{
  "message": "Explain this timeout"
}

• 200 text/plain: assistant reply (includes fallback when n8n is unavailable)
• 400 application/json: invalid input
• 500 application/json: unexpected internal error

GET /health

{ "ok": true }

GET /api/agent/metrics

Returns real-time dashboard metrics from Supabase (or in-memory runtime fallback when unavailable).

{
  "ok": true,
  "totalExecutions": 12842,
  "criticalErrors": 12,
  "avgResponseTimeMs": 248,
  "durationSampleSize": 500,
  "source": "supabase"
}

Environment Variables

PORT=3000
N8N_WEBHOOK_URL=https://your-n8n-instance/webhook/agent
SUPABASE_URL=https://.supabase.co
SUPABASE_SERVICE_ROLE_KEY=
SUPABASE_TABLE=agent_messages

# n8n retry policy (429)
N8N_RETRY_MAX=2
N8N_RETRY_BASE_DELAY_MS=800

# API rate limit (per IP)
RATE_LIMIT_WINDOW_MS=60000
RATE_LIMIT_MAX_REQUESTS=20
RATE_LIMIT_BLOCK_MS=30000

# Metrics sampling
METRICS_SAMPLE_LIMIT=500

Local Setup

1. Install dependencies: npm install
2. Configure .env from .env.example
3. Run app: npm run dev
4. Open http://localhost:3000

Supabase Table

create table if not exists public.agent_messages (
  id bigserial primary key,
  created_at timestamptz not null default now(),
  message text,
  response_body jsonb,
  ok boolean not null default false,
  error_message text
);

Troubleshooting

• 429 from /api/agent: request burst hit IP rate limit; wait for Retry-After.
• n8n returns 429: tune N8N_RETRY_MAX and N8N_RETRY_BASE_DELAY_MS.
• 500 on /api/agent: validate N8N_WEBHOOK_URL and webhook execution in n8n.
• No saved history: check Supabase credentials (SUPABASE_URL and service key).
• Metrics unavailable: check Supabase access; endpoint automatically falls back to runtime counters.
• Render boot issues: confirm npm start and Node version >= 18.

Recent Changes

• v1.3.1: final UI/docs polish with integrated header, refined About section, demo in docs and version consistency across app/documentation.
• v1.3.0: dashboard connected to real metrics via /api/agent/metrics.
• v1.2.1: retry with exponential backoff for n8n 429 responses + API rate limit per IP.
• v1.2.0: docs and interface reorganized into Chat, Changelog and About Project sections.
• v1.1.0: interaction persistence migrated to Supabase Postgres.